Optimal volume backscattering thresholds for echo integration

Abstract

Echo integration is a commonly used technique for assessment of fish stocks. In echo integration, a frequently used method to increase the signal-to-noise ratio is to ignore data below a set volume backscattering threshold. Although this form of thresholding is common, objective and replicable techniques for choosing this threshold are rarely used. Two opposing goals come into play when choosing an optimal threshold for estimating fish biomass using echo integration: maximizing the energy from backscatterers of interest, while simultaneously minimizing the energy from backscatterers not of interest. Both empirical and modeling techniques for choosing optimal thresholds were demonstrated using data from the 2003 integrated acoustic and trawl survey of Pacific hake. Empirical techniques were based on the receiver operating characteristic (ROC) curve. An ROC curve is a graphical plot of the number of true positives versus false positives for a binary classification system as the discrimination threshold is varied. Modeling techniques were based on volume backscattering data generated from modeled TS and density of hake and nonhake scatterers. For the case study using the 2003 Pacific hake survey, the historical threshold used for the southern portion of the survey was shown to be nonoptimal for that year.